ST-AGRNN: A Spatio-Temporal Attention-Gated Recurrent Neural Network for Traffic State Forecasting

Table 3

Comparison of traffic speed prediction results of the ST-AGRRG and ST-DWGRU models (bold is the best).

T	Metric	PeMSD4		PeMSD8
		ST-AGRNN	ST-DWGRU	ST-AGRNN	ST-DWGRU
15 min	MAE	1.19	1.20	1.015	1.005
	RMSE	2.36	2.40	2.07	2.08
	MAPE	2.17	2.21	1.82	1.81
30 min	MAE	1.45	1.48	1.24	1.25
	RMSE	2.98	3.12	2.63	2.70
	MAPE	2.69	2.75	2.21	2.24
60 min	MAE	1.76	1.90	1.53	1.57
	RMSE	3.63	4.01	3.33	3.49
	MAPE	3.24	3.53	2.71	2.78